On 06/23/2015 08:23 PM, Jakub Hrozek wrote:
On Tue, Jun 23, 2015 at 05:00:27PM +0200, Lukas Slebodnik wrote:
> On (17/06/15 20:44), Lukas Slebodnik wrote:
>> On (17/06/15 15:32), Lukas Slebodnik wrote:
>>> ehlo,
>>>
>>> I was able to reproduce failed build in covergae phase on local machine.
>>>
>>> It seems that gcc did some optimization and
>>> used execve instead of execle when the code was
>>> instrumented for coverage analysis.
>>>
>>> So the exec* function was not wrapped and it tried to call
>>> real binary ipa-getkeytab. Obviously it failed.
>>>
>>> LS
>>
>> >From 727d2dae8b65697269bd2e43adeb06500d65ac11 Mon Sep 17 00:00:00 2001
>>> From: Lukas Slebodnik <lslebodn(a)redhat.com>
>>> Date: Wed, 17 Jun 2015 14:29:09 +0200
>>> Subject: [PATCH] test_ipa_subdomains_server: Fix build with --coverage
>>>
>>> It seems that gcc did some optimization and
>>> used execve instead of execle when the code was
>>> instrumented for coverage analysis.
>>>
>>> So the exec* function was not wrapped and it tried to call
>>> real binary ipa-getkeytab
>>
>> It still fails on rhel6
>>
> A crash/fail in rehl6 is unrelated to my patch.
> make check will fail on different place.
>
> So this patch can be revied and if somebody have free cycles
> he can look into the new problem on rhel6.
I don't therefore we need a ticket I guess, otherwise we can't push
Nikolai's patch.
Lukas's patch fixes the issue on fedora rawhide and debian testing.
The code looks good and harmless (just added one wrapper).
We have ticket to track the issue on rhel6 which has different root
cause.
So ACK from me to Lukas's patch.
Link to CI without Lukas's patch:
http://sssd-ci.duckdns.org/logs/job/18/20/summary.html
With Lukas's patch:
http://sssd-ci.duckdns.org/logs/job/18/19/summary.html
Michal